Communications of the ACM
Harnessing Spammers to Solve AI Problems
Computers are poor at correctly orientating images like this one, but were the problem to be built into Web site security, spammers would soon start working hard to solve it.
Tatters / http://www.flickr.com/photos/tgerus/
Some Completely Automated Public Turing Test To Tell Computers and Humans Apart (CAPTCHAs) security systems are already being solved by spammers, but CAPTCHA co-creator Luis von Ahn says that when a software program is written that can solve all CAPTCHAs it will be a cause for celebration as well as concern. He says that when spammers are finally able to break any CAPTCHA it will mark a major milestone in artificial intelligence (AI).
Von Ahn says he has seen offers as high as $500,000 for anyone capable of writing software that can break reCAPTCHA, an improved system used by several major Web sites. The $500,000 prize is enough to attract people with the skills to accomplish such a feat, and is five times bigger than the Loebner Grand Prize that is offered to the programmer who designs a computer that passes the Turing test. Von Ahn says if spammers are able to write a program that reads distorted text they will have solved an AI problem, and security groups will be able to switch to alternative CAPTCHA systems, such as one based on pictures, which will give spammers another AI problem to solve.
For example, a system Google will present at the International World Wide Web Conference asks users to correctly orient images that are randomly rotated, a task computers struggle with when an obvious horizon is unavailable. The ability to easily change security measures could make using spammers as an AI resource tool a viable option.
From New Scientist
View Full Article
Abstracts Copyright © 2009 Information Inc., Bethesda, Maryland, USA 
No entries found